Multimodal Aspect-Based Sentiment Analysis (MABSA) involves extracting aspect terms from text-image pairs and identifying their sentiments. Most existing tasks consider one fixed sentiment category with explicitly mentioned aspects. However, these tasks seldom consider expressive sentiment categories, implicit aspects, and explainability. To this end, we introduce a novel task of Open-domain Explainable Multimodal Aspect-Based Sentiment Reasoning (OX-MABSR). This task enables the prediction of open-vocabulary aspect-sentiment pairs, together with the generation of sentiment explanations and reasoning paths. To benchmark OX-MABSR task, we construct OX-MABSR-Bench, a dataset annotated with explicit and implicit aspects, expressive sentiment categories, as well as perceptual and cognitive two-level explanations. The explanations capture visual and textual cues, including aesthetics, facial expressions, scenes, and textual semantics, together with background and situational knowledge. In addition, we annotate the reasoning paths that trace how the sentiment evolves from surface cues to a deeper contextual understanding. To address OX-MABSR task, we propose MABSR-LLM. Extensive experimental results show our MABSR-LLM outperforms strong baselines. To the best of our knowledge, we are the first to provide a unified framework for open-domain and explainable MABSR.

Recent studies on textual Aspect-Based Sentiment Analysis have utilized graph neural networks with dependency trees and self-attention respectively, to learn the information of syntactic structures and semantic correlations of the given sentence. However, these methods failed to address issues of syntactic insufficiency caused by the dependency parsing capability and imperfect attention weights calculated through self-attention with hidden representations. In this paper, we propose a sentiment-aware dual-flow graph convolutional networks (SeDuFo) model using commonsense knowledge. To be specific, in KSyn-Flow, to encourage the importance of words conveying sentiment, sentiment values from SenticNet and part of speech, are used to enhance the matrix constructed from the dependency tree. Mean-while, in KSem-Flow, to better capture semantic correlations between words, word embeddings trained from ConceptNet are leveraged to enhance the representations with more structured prior knowledge. Experimental results on three public datasets verify the effectiveness of our proposed model, which outperforms the state-of-the-art methods.